Asymmetry of the exchange rate pass-through:
An exercise on the Polish data1
Jan Przystupa2
Ewa Wróbel3
Abstract
We propose a complex investigation of the exchange rate pass-through in a small open economy in transition.
We assess the level, linearity and symmetry of exchange rate pass-through to import and consumer prices in
Poland and discuss its implications for the monetary policy.
We show that the pass-through is incomplete even in the long run. There is pricing to market behaviour both in
the long and short run. We do not find a strong evidence of non-linearity in import prices reaction to the
exchange rate and reject the hypothesis of an asymmetric response to appreciations and depreciations.
On the other hand, we find an asymmetry of CPI responses to the output gap, direction and size of the exchange
rate changes and to the magnitude of the exchange rate volatility. The asymmetry is mostly visible after
exogenous shocks. We reject the hypothesis of an asymmetric reaction of prices in a high and low inflation
environment. 4
JEL Classification: C22, E52, F31.
Key Words: Exchange Rate Pass-through, Non-linear Model.
1
We are thankful to Michael Frömmel, Ryszard Kokoszczyński and Tomasz Łyziak for many helpful
comments.
2
National Bank of Poland and Institute for Market, Consumption and Business Cycles Research, Warsaw;
corresponding author: Jan.Przystupa@nbp.pl
3
National Bank of Poland.
4
The usual disclaimer applies.
1. Introduction (non-technical summary)
Understanding how prices respond to exchange rate is of key importance for any small open
economy, in particular, for the economies in transition with a free floating exchange rate.
Poland is an example of such country. In this paper we show the main properties of the
exchange rate pass-through in Poland and assess their implications for the monetary policy.
Our analysis covers a period 1997-2008, which is characterized by a homogenous monetary
policy (inflation targeting, flexible exchange rate) and a generally falling inflation rate. We
assess the degree of pass-through to import prices, examine if they exhibit pricing to market
behaviour and whether they react to the exchange rate symmetrically. We find that pricing to
market exists, both in the long run and in the short run, but reject the hypothesis of
asymmetric reaction to the exchange rate. Then, we pass to the CPI, show the level of passthrough and examine whether the process is symmetric. In contrast to import prices, in the
case of CPI we find some asymmetries with respect to the output gap, direction and size of
exchange rate change and its variability. We do not find asymmetry regarding inflation
environment. Finally, we discuss implications for the monetary policy. To our knowledge,
both pricing to market and asymmetric reactions of import prices and CPI have not been
explored on the Polish data.
The existing Polish literature on the pass-through is not particularly abundant. It is mostly
based on McCarthy (2000) where the impact of a sequence of supply, demand and exchange
rate shocks on the import, producer and consumer prices is examined, see Przystupa (2002),
Cholewiński (2008). Since it is a popular method, for the sake of comparability with other
studies, we also provide results we have obtained this way. They are shown in tables A1-A4
in the appendix. Generally, pass-through of the exchange rate changes to the import prices do
not vary over time and slightly exceeds 0.7 after 8 quarters. On the other hand, pass-through
to the consumer prices exhibits a notable fall: it dropped by half over the period 2002-2008
from 0.42 to 0.21. Such high pass-through effect at the earlier stages of transformation can be
interpreted as a result of fast structural changes in the economy, while the huge drop of the
pass-through in the last years as an ‘internationalization’ of the Polish economy due to a
massive inflow of the export-oriented foreign direct investment.
Figures for the pass-through effect to the import prices are close to results for other developed
economies; whereas figures for the consumer prices differ considerably (they are close to the
results estimated for 1971-1983 for the countries with the direct inflation targeting, see Ihrig
et al. (2006).
Pass-through to the import prices, estimated with other methods gives similar results - we can
safely say that the long-run pass-through for both nominal effective exchange rate and the
bilateral exchange rate lies within the range of 0.7-0.8; moreover, a strong evidence of pricing
to market is confirmed. Estimating a dynamic equation for import prices we show that
instantaneous pass-through is 0.51 – much lower than in the long run and there is weaker
evidence of pricing to market. Also, the error correction works quite efficiently: 46% of
disequilibrium is eliminated within one quarter. Checking the nonlinearity of import prices
reaction to appreciations and depreciations we find that if there is appreciation of the Polish
currency with respect to the euro then the effect of the instantaneous pass-through is 0.55 and
there is pricing to market. 65% of disequilibrium is eliminated within one quarter. A different
picture emerges for the depreciation periods: neither pricing to market nor error correction
mechanism seems to operate. The only factor affecting import prices is the exchange rate. The
pass-through effect is 0.59. We cannot reject the hypothesis that pass-through in depreciations
2
and appreciations is equal and tentatively conclude therefore, that there is no asymmetry in
import prices’ reaction to the exchange rate.
For the CPI we found the strong asymmetry of the pass-through effect along the business
cycle: this effect may exceed 0.27 in the early expansion, and then it drops to 0.18-0.19 in the
peak and to around zero in the early recession. During the late recession and the trough it is
growing up to 0.18-0.19 and to over 0.27 in the early expansion. This is coherent with the
behaviour of enterprises in the business cycle, conditioning their investment decisions on
expected profits with a maximum in the early expansion and a minimum in the early
recession.
On the other hand, a strong asymmetric reaction was identified in the periods of appreciations
and depreciations. If the zloty appreciates, the pass-through to the consumer prices declines
and varies from 0.02 to 0.07. During the depreciations or slight appreciations, the passthrough effect equals 0.24. Also regarding volatility the pass-through effect seems to be
strongly asymmetric: for the volatility over 4.3% it is equal to 0.25, and 0.55 otherwise. The
lower pass-through in case of higher volatility can reflect the producers’ reluctance to
frequent price changes due to the menu costs.
Examining the role of the inflation environment in the exchange rate pass-through behaviour
we do not obtain a clear picture, hence we decide to reject the hypothesis that the asymmetry
exists.
2. Brief review of the literature
Theoretical exchange rate pass-through literature is dominated by papers dealing with the
degree of price adjustment and factors that affect pass-through. The discussion started after
numerous tests on variety of goods and across countries yielded very little evidence
supporting an assumption of the absolute or relative Purchasing Power Parity (Rogoff (1996),
Ihrig et al. (2006)..
One strand of the literature investigates pass-through stressing the role of market organization,
product segmentation, pricing to market and competition as factors explaining why it is
incomplete. In a seminal paper, Dornbusch (1987) points out that firms operating under
imperfect competition may adjust mark-ups in response to the exchange rate shock. Froot and
Klemperer (1989) show that if market share matters for future profits, then firms facing an
exchange rate appreciation will decide whether to raise current or future profits, depending on
their perception of durability of the appreciation. This strand of literature is closely related to
the problem of pricing to market behaviour and we shall characterize it in more details in
section 2, where we describe the model of import prices.
In the second strand of this literature, pass-through is treated as a phenomenon affected by the
monetary policy. In a seminal paper Taylor (2000) suggests that price responsiveness to the
exchange rate is positively related to the inflation rate. Bailliu and Fujii (2004) find that a
transition to a low inflation environment induced by a shift in monetary policy is a cause of a
declining pass-through. Devereux and Yetman (2002) build a model in which sticky prices
result from menu cost of price adjustment. Monetary policy determines the average rate of
inflation and exchange rate volatility. Firms are allowed to choose frequency of price
adjustment and optimal frequency varies with the monetary policy regime. For a given menu
cost, the higher the rate of inflation and the more volatile the exchange rate, the higher the
3
frequency of price adjustment decisions taken by firms and the higher the pass-through effect.
As a result Devereux and Yetman argue, that pass-through is endogenous to monetary policy.
Devereux, Engel and Storgaard (2003) analyze the determinants of an exporting firm’s choice
of currency in which it pre-sets prices. With nominal price stickiness, the aggregate degree of
exchange rate pass-through is determined by this decision. They develop a model of an
endogenous exchange rate pass-through, in a framework in which the exchange rate itself is
also endogenously determined. They find a two-sided relationship, namely that exchange rate
volatility determines the price setting behaviour of a firm, and therefore the degree of the
pass-through, and that the degree of the pass-through determines the volatility of an exchange
rate. On the other hand, Campa and Goldberg (2002) show that higher inflation and exchange
rate volatility are weakly associated with higher pass-through.
To sum up this part of the literature overview, factors that explain an incomplete pass-through
in the long run can be considered microeconomic, whereas these affecting its degree in the
short run – macroeconomic.
Linearity and possible asymmetry are investigated less frequently, even though a question
whether they exist is important if not crucial for the monetary authorities. Linear response
means that prices react in the same way to “big” and “small” changes in the exchange rate.
Many papers rather assumed linearity than tested it. But if there are some menu costs of price
adjustment, then relatively small changes in the exchange rate may not be passed onto prices.
Only after exceeding a certain threshold these changes might be passed on customers. On the
other hand, if domestic prices react more to, say, depreciations than to appreciations, meaning
there is an asymmetry in the price reaction to the exchange rate changes, it may reflect
distortions in the markets, weak competition in particular.
Bussière (2007) provides assumptions and discusses the economic meaning of both quadratic
and cubic non-linearities. The quadratic case relies on three assumptions: (i) that export
prices are rigid downwards (ii) that quantities of exported goods are rigid upwards (exporting
firms work at full capacities) and (iii) that exporting firms care about their market share.
Exporters try to offset a fraction of exchange rate movements, and this fraction varies with the
magnitude of the exchange rate changes. For a larger appreciation, the reaction of export
prices decreases, whereas for a large depreciation – increases. Also, in appreciations exporters
reduce their prices by a larger amount, because they fear losing market share. The cubic case
relies on the assumption of the existence of menu costs on the side of both exporters and
importers. If the former is true, then exporters only slightly adjust prices in their currency in
response to the exchange rate changes, while these changes are reflected in the import prices.
If the menu costs are at the side of importers, the exporting firms adjust prices to offset small
changes in the exchange rate.
Recently the gap between empirical literature assuming linearity and symmetry and that
testing it has been shrinking. Early works on this topic are Mann (1986) and Goldberg (1995);
both find asymmetries in consumer prices’ reaction to appreciations and depreciations. More
papers appeared only after the year 2000. Devereux and Yetman in the paper mentioned
above use a cross-section time series model to show that there is a non-linear relationship
between the degree of pass-through and inflation rate. As inflation rises, pass-through
increases too but at a decelerating rate. Pollard and Coughlin (2003) examine the symmetry in
response of import prices in manufacturing to appreciations and depreciations as well as to the
size of the change of the exchange rate in the US and show that there is some asymmetry in
4
response of import prices in different industries. They find no clear direction of the
asymmetry across industries, however, and conclude that pass-through is positively related to
the size of the change in the exchange rate. Campa et al. (2006) examine non-linearity in
import price adjustment in the EU (15). They explore three types of non-linearities:
adjustments which are not proportional to the size of the deviation, adjustments which are
non-symmetrical to the sign of the deviation and finally thresholds below which no
adjustment takes place. Using data on the industry-level they obtain a strong evidence for the
presence of nonlinearities in the adjustment towards long-run equilibrium in certain
industries. They also find evidence for the existence of thresholds of no adjustment centred on
zero. The thresholds seem to be much lower for the manufacturing than for the commodity
industries. Herzberg et al. (2003) use aggregate data for the UK. They do not find evidence of
non-linearity in the reaction of import prices. Bussière (2007) investigates whether export and
import prices of G7 countries exhibit linear and symmetrical reaction to the exchange rate
movements. He finds that a non-linear effect cannot be neglected, although the direction of
the asymmetries and the magnitude of non-linearities vary across countries. Wickremasinghe
and Silvapulle (2004) find that in Japan import prices of manufacturing display a statistically
significant difference in their adjustment to appreciations and depreciations and that the
reaction to the former is bigger than to the latter.
There is less evidence of import price reactions in the emerging markets. Alvarez et al. (2008)
report a weak asymmetry between the reaction of import prices to appreciations and
depreciations in Chile. They examine unit import values and wholesale import prices, the
former aggregated into three subcomponents by end use: consumer, intermediate and capital
goods; the latter distinguished imported goods from three sectors – mining, agriculture and
industry. They find that coefficients for depreciations are higher than for appreciations, but
null hypothesis that both coefficients are equal is not rejected only marginally for capital
goods; for agriculture only depreciations are passed into import prices.
Literature dealing with the consumer price index is even scarcer. Goldfajn and Werlang
(2000) use a panel of 71 countries and report an asymmetric reaction of pass-through over the
business cycle. Correa and Minella (2006) investigate non-linearities in pass-through to
consumer prices in Brazil. They estimate a Phillips curve with a threshold for the passthrough and show that the magnitude of a short run pass-through is higher when the economy
is growing faster, when the exchange rate depreciates above a certain threshold and –
surprisingly – when exchange rate volatility is lower. Posedel and Tica (2007) use a TAR
model on Croatian data and find a single threshold of a monthly change in the exchange rate
below which it does not affect consumer price inflation. Khundrakpam (2007) finds
asymmetry of pass-through related to the direction of the exchange rate changes in India.
To sum up, there is growing evidence showing the existence of various types of nonlinearities both on the first stage of the pass-through, i.e. import prices, as well as on the last
stage – i.e. consumer prices.
Our paper fits in this type of literature. The first part is devoted to the analysis of import
prices: we start with a linear equation, and afterwards test whether it is mis-specified, i.e.
whether it lacks non-linear terms. We do not find strong evidence for non-linearities, but it
seems that in depreciations and appreciations different mechanisms of adjustment operate. In
appreciations of the home currency with respect to the euro pricing to market seems to have
some impact on import prices. However, this impact disappears in depreciations. The next
part of the paper is focused on retail prices. We estimate threshold models and models based
5
on nonreversibility of the linear functions and find an asymmetric reaction of consumer prices
to appreciations and depreciations, to the volatility of the exchange rate, and, in the models
based on nonreversibility, to the business cycle fluctuations. The impact of the inflation
environment does not show any apparent asymmetry.
3. Import price model
There is a vast literature pointing that import prices can be affected by domestic prices. The
phenomenon is known as pricing to market (henceforth PTM). It was perceived as a result of
market imperfections (monopolistic power of price setters) and market segmentation, since in
this case producers can charge a mark-up on costs. The mark-up depends on the elasticity of
demand for a given product and this in turn is determined by competitors’ prices. Facing a
change in the exchange rate, producers can decide whether and to what degree the mark-up
should absorb these changes. For example, if the domestic currency of an exporting firm
depreciates, it can refrain from increasing its mark-up to build a future market share. Firms
having a power to control their prices can charge different prices in different markets. Pricing
to market can therefore explain why pass-through tends to be incomplete.
Further analyses of PTM show however, that imperfect competition together with market
segmentation is not a sufficient condition for a persistently low degree of pass-through5. In
this vein, Bergin and Feenstra (2001) use a general equilibrium model to show that segmented
markets with identical preferences cannot explain incomplete pass-through in contrast to
segmented markets with different preferences. Therefore, they consider translog preferences
in which the elasticity of demand varies with the price a firm sets. Corsetti and Dedola (2002)
analyze market segmentation resulting from a vertical interaction between monopolistic
producers and retailers. They showed that due to the presence of downstream retailers,
upstream firms with monopoly power may face different demand elasticities in national
markets even under symmetric constant elasticity preferences across countries. Monopolistic
firms optimally set different prices to domestic and foreign dealers, and therefore the law of
one price fails to hold at producer and consumer levels.
In this paper we use a model which is common in the literature on pass-through effect (e.g.
Goldberg, Knetter (1997), Bache (2002), Bahroumi (2006). We assume that an exporting firm
has some control over its price in the importing country. Namely, it sets the price as a markup over its marginal cost. In the importing country, import price expressed in its own currency
is equal to the exporting firm’s price multiplied by the exchange rate. Thus:
(1) Pt F = λt CtF ,
(2) Pt IMP = Et Pt F = Et λt CtF ,
where Pt F represents export prices of a foreign country, CtF is a marginal cost of production
in a foreign country, λt is a firm’s mark-up in the importing country, Et is the exchange rate,
and Pt IMP stands for import prices of a home (importing) country.
5
For a more detailed discussion see for example Herzberg et al. (2003), or Bache (2007).
6
Mark-up depends on the competitive pressure from importer’s domestic production and
demand pressure, Yt . Competitive pressure can be expressed as a relation between domestic
prices and foreign prices in terms of importer’s currency:
α
 Pt H 
(Yt H ) β
(3) λt = K 
F 
 Et Ct 
Combining (2) and (3) gives:
(4) Pt IMP = K ( Et CtF )
1−α
(P )
H α
t
0 < α < 1, 0 < β < 1 and K is constant.
(Yt H ) β ,
taking natural logs (depicted in lower case) we have:
(5) ptIMP = κ + (1 − α ) et + (1 − α ) ctF + α ptH + β ytH .
In practice it is often difficult to obtain in the empirical study equal parameters at the
exchange rate and foreign prices that usually serve as a proxy for costs in the exporting
country. Bache (2002) and Barhoumi (2006) stress that exchange rates are more volatile than
costs, therefore the extent to which they are passed on prices may differ. One usually starts
estimations without this restriction and tests it in the next step. Thus, the estimated equation
is:
(6) ptIMP = α 0 + α1et + α 2 ctF + α 3 ptH + α 4 ytH + ε t .
We shall proceed this way. Some authors also point that domestic demand often turns out to
be statistically insignificant (see for example Bussière (2007)). The rationale for that is
simple: information that is contained in the measure of demand pressure can be already
incorporated in another variable, notably in domestic prices. This is why this term is
frequently omitted in estimations. To verify whether it is justified to allow for the output
pressure in the dynamic (short-run) equation, we have checked the correlation between the
output gap measure (a difference between actual and HP-filtered GDP) and the change in
domestic prices – it was quite high, namely 0.538, therefore we decided not to plug the
output gap into the equation.
3.1. Estimation method and data
Many studies of the pass-through effect are based on short-run dynamic equations. Such
estimate has also been made for Poland (Przystupa (2002), Cholewiński (2008)). In this paper
we present estimates of a long-run reaction of import prices to the exchange rate. If variables
are cointegrated, as they seem to be in this case, then limiting the analysis to the short-run
would reduce the information content. On the other hand, cointegration analysis requires data
span that is sufficient for inferences on equilibrium levels of the variables. Our analysis
covers an eleven-year period only (1997.Q1 2008.Q1). It is relatively short, which makes our
conclusions somewhat risky. It needs to be stressed that estimation of asymmetry of exchange
rate pass-through relies on an even smaller number of observations, since our sample must be
divided into two sub-samples that reflect two different states, e.g. appreciation and
depreciation or “big” and “small” changes of the exchange rate. Our results should therefore
be treated as guidelines rather than sharp point estimates.
7
Taking into account the shortness of our sample and the small number of usable observations,
we decided not to confine the analysis to one method only, but – whenever possible – to use
more to reduce uncertainty concerning robustness of our results. Thus, in the case of import
prices we use Johansen cointegration method and fully modified least squares. The latter is
particularly well suited if there are problems with endogeneity and serial correlation.
It is important to note that the period covered in this paper is relatively homogenous in terms
of a monetary policy regime, namely inflation targeting and flexible exchange rate. This
makes the estimation easier and more reliable and therefore partially offsets the drawbacks of
the time span shortness. On the other hand, the EU entry in 2004 constituted a significant
disturbance to the real economy. There was an increase in inflation, but also – what is more
important for this study – the role of exchange rate as a variable affecting the real sector of the
economy has started to change since then. Due to an increasing role of intra-company trade,
the impact of the exchange rate on the aggregate demand diminished (see also Łyziak et al.
(2008) and Przystupa and Wróbel (2006)6. Moreover, Poland’s foreign exchange market
started to be increasingly impacted by the euro, whereas the position of the US dollar was
falling.
We begin our estimations with the nominal effective exchange rate, but then restrict our
discussion to the import prices’ behaviour with respect to the bilateral exchange rate
(EUR/PLN7). One reason is the big share of imports from the EU in the total imports (about
two thirds in 2007), another is the fact that imports from the euro area consist mostly of
manufactured products. Pricing to market and asymmetries in the pass-through, i.e. the issues
that are in the centre of our interest, are a feature of manufacturing rather than raw materials.
Since our aim is to check whether the exchange rate pass-through in Poland is linear and
symmetric, we start with a linear equation and then test whether quadratic and cubic terms are
lacking in the dynamic equation. Then we check whether import prices react asymmetrically
to appreciations and depreciations.
In the estimation we use quarterly data. They are seasonally adjusted with X11. The only
exception is the nominal exchange rate, which is not adjusted since, in the case of Poland, it is
not significantly affected by seasonal factors. To assess the role of PTM, we use Producer
Price Index (source: Main Statistical Office (GUS)); due to the data problems till the end of
1999 it is an overall index, whereas since 2000 it comprises goods produced for the domestic
market only. As shown by graph A5 in the Appendix, the overall index and the index of prices
for domestic market started to diverge mostly in 2004, thus we assume that this should not
significantly influence our results. As a proxy for foreign prices we use euro area export
prices (source: Eurostat). We do not employ unit export values of Poland’s main trading
partners. Silver (2007) points that unit export values can be substantially biased in
representing export and import prices. Import price data come from GUS, whereas nominal
exchange rate (quarterly average) - from the National Bank of Poland. All data are in logs, the
base year for price indices is 2000.
As it is clear from table A9 foreign and domestic prices as well as the exchange rate are non
stationary. One may have some doubts whether import prices are stationary. ADF test shows
that seasonally adjusted data with an intercept can be considered as I(0), whereas raw data are
6
The authors show changes in the exchange rate regime and the role of the exchange rate in the monetary
transmission.
7
Upside movement of the exchange rate means appreciation.
8
rather non-stationary. Johansen test (Table A5 – A7) indicates that there is one cointegrating
relation, therefore we treat import prices as an I(1) variable8.
3.2. Estimation results
As we have mentioned above, there is one cointegrating equation of import prices, domestic
prices, foreign prices and the exchange rate. Both nominal effective exchange rate (NEER)
and the bilateral exchange rate give the cointegration. The cointegrating vector is presented in
tables A8 and A9. We start estimates with the unrestricted vector – all variables have a proper
sign. Contrary to domestic prices, foreign prices are not significant. Significance of domestic
prices means that there is pricing to market behaviour in the long run. The long-run exchange
rate pass-through is 0.91 for the nominal effective exchange rate and 0.71 for the bilateral
exchange rate.
Now we pass to the restricted version of the cointegrating relation. First we check whether the
coefficients at NEER (or the bilateral exchange rate) and coefficients at foreign prices are
equal. In both cases a Chi-square test shows that we cannot reject this hypothesis. In the
restricted version the pass-through is approximately 0.8 for both NEER and the bilateral
exchange rate. All variables are statistically significant, and the significance of domestic
prices implies the existence of a pricing to market effect. Next, we check whether the passthrough effect is full. This hypothesis is not rejected at the 1% level for the bilateral exchange
rate but rejected for the nominal effective exchange rate. Neither the restriction α1 = α 2 = 1,
α 3 =0 (the law of one price) nor the restrictions α1 = α 2 , α 2 + α 3 =1 (unit homogeneity) and
α1 = α 2 =1 (unit coefficients on the exchange rate and foreign prices) for the bilateral
exchange rate can be rejected at the 1% level. For the nominal effective exchange rate only
the law of one price cannot be rejected at the 1% level. All these restrictions, with exception
of the law of one price for the bilateral exchange rate are rejected at 5% significance level;
this however can be rejected at the 10% level. All in all, we conclude that while for the
nominal effective exchange rate the results seem quite clear, for the EUR/PLN exchange rate
we need a cross-check.
To have a clearer picture we estimate the model with FMLS and then impose the same set of
restrictions. Table A9 (right hand panel) shows that, with the exception of foreign prices, in
the unrestricted cointegrating vector all parameters are statistically significant, and all
coefficients obtained with these two methods are similar. The point estimate of long-run passthrough is 0.68. Data show that there is long-run pricing to market. As previously, restriction
α1 = α 2 cannot be rejected. Once again, all parameters are statistically significant. Contrary
to the results obtained with the Johansen method, all remaining restrictions are rejected at the
1% level of significance.
To sum up this part of our estimates, we can safely say that the long-run pass-through for both
the nominal effective exchange rate and the bilateral exchange rate lies within the range of
0.7-0.8, and that there is equality of the coefficients of the exchange rate and foreign prices.
There is no strong support in the data for the existence of the law of one price, unit long-run
coefficients on the exchange rate and foreign prices, and long-run unit homogeneity in
8
Hjalmarsson and Österholm (2007) show that Johansen’s trace test and maximum eigenvalue test may bring
erroneous results under the situation of nearly integrated variables. They suggest performing additional test
which relies on a system of restrictions imposed on the cointegrating vector. We have performed this test and do
not reject the hypothesis that the variables are cointegrated in the long run.
9
domestic and foreign prices. A final step in the cointegration analysis is the test of exogeneity.
All tested variables - exchange rate, foreign prices and domestic prices - seem to be weakly
exogenous. (Table A10). Exchange rate is strongly exogenous, but domestic producer prices
are not9.
Then we estimate a dynamic equation for import prices using cointegrating vector obtained
from the FMLS, namely:
4
4
4
4
i =0
i =0
i =0
i =1
(7) ∆ptIMP = β 0 + β1 ECt −1 + ∑ β 2,i ∆et −i + ∑ β 3,i ∆ptF−i + ∑ β 4,i ∆ptH−i + ∑ β 5,i ∆ptIMP
− i + vt ,
where EC is the error correction term and ∆ stands for the first difference of a variable.
After elimination of statistically insignificant parameters we get the following results (Table
A11): instantaneous pass-through is 0.51 – much lower than in the long run, there is weaker
evidence of pricing to market in the short run than in the long-run. On the other hand, the
error correction works quite efficiently: 46% of disequilibrium is eliminated within one
quarter.
To check whether import prices react in a non-linear way to the exchange rate we perform a
test similar to the Ramsey’s RESET test, but with the Taylor expansion for the non-linear
function10. It should reveal whether our equation lacks square or cubic terms. There is no
strong evidence of the lacking square terms (Statistic F(1.39) =0.267 sign. level 0.61,
LM(1)=0.299 sign. level 0.58). However, it is possible that there is a very weak cubic nonlinearity (F(2,38) =2.001 sign. level 0.15, LM(2)=4.19 sign. level 0.12).
Our next step is to examine how changes in the exchange rate are transmitted to the import
prices in the periods of appreciation and depreciation of the Polish currency. Thus we set:
At = {
1 if ∆ e t ≺ 0
0 otherwise
and
Dt = {
1 if ∆ e t 0
0 otherwise
and in equation (7) we replace ∆et with β1A ( At ∆et ) + β1D ( Dt ∆et ) and estimate separately the
equation for the appreciation and depreciation of the exchange rate.
The results are reproduced in tables A12 and A13. If there is appreciation of the Polish
currency with respect to the euro the correcting mechanism is very efficient: 65% of
disequilibrium is eliminated within one quarter. There is pricing to market, which means that
exporters in the euro zone do not increase mark-ups, but rather care about future market share.
The effect of the instantaneous pass-through is 0.55, but it is estimated with a low T-statistic.
A different picture emerges for the depreciation periods: Neither pricing to market nor error
correction mechanism seems to operate. The only factor affecting import prices is the
exchange rate. The pass-through effect is 0.599. We could not reject the hypothesis that passthrough in depreciations and appreciations is equal. We tentatively conclude therefore, that
there is no asymmetry in import prices’ reaction to the exchange rate. Pricing to market
manifests itself in periods of appreciations of the domestic currency only. When exporters’
currency appreciates the exporters do not tend to lower their mark-ups. What can be the
reason for such asymmetry? There are at least three, seemingly consistent, arguments. First,
exporters expect appreciation rather than depreciation of the Polish zloty with respect to the
9
We do reproduce the results of Granger causality here.
The same test is used in Herzberg et al. (2003)
10
10
euro and treat depreciations as short-lived and therefore are not afraid of losing market share.
Second, prices are rigid downwards, exporters therefore are not eager to adjust them if their
currency appreciates. Third, if exporters do not increase their mark-ups in favourable
conditions, then they would incur losses if they decided to lower them in a case of the
appreciation of their currency.
We have also checked how pass-through works in two additional cases: a positive and a
negative sign of the error correction term. The former means that import prices are higher
than the equilibrium level determined by exporters’ prices and domestic prices due to, for
example, too high transportation costs, inclusion of other costs in the price (like insurance or
hedging costs), or misperceived intensity of competition. If this is the case, a process of return
to the equilibrium relies mostly on exchange rate adjustment. There is also statistically
significant pricing to market behaviour, but its impact is very limited. Depreciation of the
domestic currency exchange rate is a factor that increases other prices (domestic prices).
Pricing to market is small probably because in this case importers can involve only a small
part of transportation and distribution costs in their prices. If import prices are lower than the
equilibrium level, then the error correcting mechanism works quickly and efficiently
eliminating 65% of the disequilibrium in one quarter; there is also a statistically significant
and relatively big impact of pricing to market, which means that distribution and
transportation costs may be overpriced.
4. Exchange rate pass-through models based on the Phillips curve.
In this chapter we examine a hybrid New-Keynesian Phillips Curve (NKPC) model. In the
hybrid NKPC for an open economy inflation is a function of three factors:
(i) next period’s expected inflation rate ( Etπ t +1 ) extended by the empirically observed
persistence of inflation (backward looking inflation π t −1 ) – e.g. Fuhrer and Moore (1995);
(ii) real marginal costs approximated by the output gap ( yt ) – e.g. Woodford (2003);
(iii) real exchange rate – contemporaneous ( etr ) and expected ( Et etr+1 ) – e.g. Woodford (2003).
Accepting the NKPC, we agree that behaviour of inflation depends on the slope of the Phillips
curve. If the Phillips curve is linear, the slope is constant. . If it is non-linear, the slope is not
constant and behaviour of inflation is a function of the output gap and the exchange rate.
On the other hand, steepness of the slope of the Phillips curve depends on the determination
whether the function is convex or concave. Filardo (1998) finds that the Phillips curve is not
purely convex or concave and that convexity or concavity is determined by the output gap. If
the output gap is positive, the Phillips curve is convex (firms are more inclined to raise prices
than to lower them, but the cost of fighting inflation is falling – the slope of the Philips curve
steeping) and if the output gap is negative, the Phillips curve is concave (firms are more
reluctant to raise prices than to lower them, but the cost of fighting inflation is rising – the
slope is flatting).
In this chapter we investigate the possibility of non-linear reaction of inflation to the exchange
rate changes along the business cycle.
11
4.1. Estimation methods
As we have mentioned in the Introduction, the exchange rate pass-through to consumer prices
in Poland seems to be relatively high (about 0.2), akin to that of the developed countries in the
eighties of the last century. In such circumstances it is of key importance for the monetary
policy to know whether the process is linear and symmetric. We check pass-through linearity
with respect to (i) business cycle approximated by the output gap, (ii) appreciations and
depreciations of the nominal effective exchange rate of the zloty, (iii) volatility of the nominal
effective exchange rate and (iv) inflation environment. We distinguish two states (regimes)
for any economic process. The output gap, NEER and inflation are set up as the variables
which cause a transition from one regime to another.
We use two methods to divide the sample into properly defined sub-samples. The first one
splits the sample into classes, depending on the observation whether a given variable exceeds
some threshold. If all explanatory variables are exogenous then Threshold AutoRegresive
(TAR) models can approximate a nonlinear autoregressive structure, where the number of
regimes is small – see Tong (1990) or Hansen (2000). Caner and Hansen (2004) develop an
estimator which can be used for models with the endogenous variables and an exogenous
threshold variable. This allows estimation of thresholds for dynamic models. Basing on the
procedure and a program described in Caner and Hansen (2004) we estimate a SETAR (SelfExciting Threshold AutoRegresive) model of a general form:
Yt + θ 0 + θ1Yt −1 + … + θ pYt − p + I (Yt − d ≤ τ )(φ0 + φ1Yt −1 + … + φqYt − q ) = ε t
where:
Y – vector of explanatory variables;
d=1 and 0 ≤ q ≤ p ;
εt – white noise;
τ ∈Τ
- threshold parameter:
I(y ≤τ ) = {
1 if y ≤τ
0 otherwise
Usually the threshold value is unknown and has to be estimated. Estimation of the threshold
with the ML method needs to determine the maximum of the likelihood function, whereas the
estimation of the threshold with the TSLS or GMM method – to minimize the sum of the
squared residuals.
Our sample consists of 46 observations and due to their small number we assume that the
threshold should split the data into sub-samples with minimum 15 observations. Franses and
van Dijk (1999) show that a safe choice is min. 15% of observations in the sub-sample. In
order to estimate threshold values we use four specifications of the threshold model (based on
the Phillips curve), differing basically by the variables used as the threshold for the passthrough: output gap, NEER, NEER volatility and inflation environment.
The second method refers to Wolffram’s (1971) idea, developed further by Houck (1977), for
investigating nonreversibility of the linear functions through segmenting the variables
involved.
12
Let us assume the variable Y depends on the values of X being suspected for asymmetry, and
on the set of other variables Z. We want to examine whether one-unit increases of X from
period to period have a different impact on Y than one-unit decreases of X:
∆ Yi = a 0 +
T1
∑a
1, j
∆X
j=0
'
i− j
+
T2
∑a
2, j
j=0
∆ X i' '− j + a 3 Z i
where for i = 1,2,…,t :
∆Yi = Yi − Yi −1
∆ X i' = X i − X i −1 if Xi > Xi-1 and = 0 otherwise;
∆ X i'' = X i − X i −1
j = 1,2,…,T1,T2;
if Xi < Xi-1 and = 0 otherwise;
T1,T2 < t (j – number of lags assigned to ∆X i' , ∆X i'' ).
A nonreversibility of the function ∆Yi occurs, i.e. the reaction is asymmetric, if
T1
∑ a1, j ≠
j =0
T2
∑a
j =0
2, j
Both methods divide the sample into different sub-samples with different economic
interpretation. Output gap, suspected of the asymmetric impact, can serve as an example
(Figure 1).
Figure 1. The economic interpretation of the threshold model and model based on
nonreversibility of the linear functions
3
Nonreversible linear models
Nonreversible linear models
t1≤t0
t1>t0
2
EU accession
1
Threshold models
Threshold models
%
0
Threshold models
Russian crisis
-1
-2
19
97
Q
1
19
97
Q
3
19
98
Q
1
19
98
Q
3
19
99
Q
1
19
99
Q
3
20
00
Q
1
20
00
Q
3
20
01
Q
1
20
01
Q
3
20
02
Q
1
20
02
Q
3
20
03
Q
1
20
03
Q
3
20
04
Q
1
20
04
Q
3
20
05
Q
1
20
05
Q
3
20
06
Q
1
20
06
Q
3
20
07
Q
1
20
07
Q
3
20
08
Q
1
20
08
Q
3
-3
output gap
threshold
Source: Authors` calculations.
13
proxy of the business cycle
The output gap was calculated as a percentage deviation of the actual GDP from a potential11.
Assuming that the output gap is a proxy for the economic activity along the business cycle, its
values exceeding the estimated threshold can be interpreted as periods of prosperity (late
expansion, peak and early recession), whereas the values below the threshold – as a slump
(late recession, trough and early expansion). The threshold estimated for the output gap
identifies 25 periods of prosperity and 21 periods of slump (Figure 1).
The sub-sample derived from the nonreversibility of the linear function, for ti+1>ti, points out
periods of expansion or recovery (early expansion, late expansion and peak), while the subsample for ti+1≤ti describes periods of contraction or recession (early recession, late recession
and trough). This method identifies 24 periods of expansion and 22 periods of contraction.
The total sample contains 46 observations between 1q 1997 and 2q 2008 and seems to cover
one and a half business cycles and two exogenous shocks: “the Russian crisis” in 1998 and
Poland’s accession to the EU (2004)12.
The exchange rate pass-through to the CPI was estimated on the basis of the Philips curve
similar to the one in a highly aggregated New-Keynesian model developed at the National
Bank of Poland for inflation forecasting (see Kłos et al., 2005). The model was built on the
standard assumptions of sticky prices and wages. Basic macroeconomic relationships include:
the IS curve, uncovered interest parity (UIP) and the Phillips curve13. The shape of the
Phillips curve used to estimate the pass-through effect is as follows:
π tq,k = α1,qk Etπ t +1 + (1 − α1,qk − α 2,q k )π t −1 + α 2,q k (∆etr−1 ) + α3,q k yt −2 + ε t
i
i
i
i
i
i
where π stands for inflation (CPI),
qi is a threshold variable (i=1…4);
i=1 → output gap (y);
i=2 → ∆ nominal effective exchange rate (∆e);
i=3 → volatility of the nominal effective exchange rate (s);
i=4 → inflation (π): actual inflation – inflation target (τ);
k=1,2; k=1 for qi > τ (τ = threshold);
k=2 for qi ≤ τ
and er is a nominal effective exchange rate plus foreign inflation.
We estimate the Phillips curve using the GMM method. The instrumental variables were
derived from a set of variables explaining behaviour of the exchange rate of the zloty (see the
recalled NBP model). Nominal effective exchange rate and the output gap were also included.
Such procedure allows for relationships among the variables as in the model of inflation. The
instrumental variables were chosen in the manner assuring proper sign at all explanatory
11
As before, potential GDP was derived from the Hodrick-Prescott filter with λ=1600. To diminish the role of
the last observations, the sample was lengthen by an AR(1) process.
12
The approximate date of Poland’s accession to the EU was known since 2002 but the reaction of the economy
was hardly predictable – therefore it can be treated as a shock.
13
The output gap in the model depends on its lagged value, the ex-ante real interest rate, the real effective
exchange rate and the external demand. NEER is determined — in line with the arbitrage condition — as a
function of the USD/PLN exchange rate and the EUR/USD cross rate (exogenous variable), while the USD/PLN
exchange rate is modelled within the concept of the UIP. The explanatory variables of the USD/PLN exchange
rate are the lagged value of this variable, the interest rate disparity, the term structure of interest rates, the
EUR/USD cross rate and risk factors. The list of risk factors includes the output gap, budget deficit, net exports,
current account deficit and foreign direct investment.
14
variables, correct value of the t-statistic and p-value of the J-statistic. Finally, we estimate 16
equations with the threshold and the nonreversibility method. All details are presented in table
A15 in the Appendix. When interpreting the results, one must bear in mind a small number of
observations used in the estimation.
4.2. Estimation results.
The results seem to be consistent with these obtained by Mann (1986), Goldberg (1995),
Goldfajn and Verlang (2000), and Correa and Minella (2006). They are presented in Table 1
and Figure 2.
Table 1. The asymmetry of the exchange rate pass-through to the consumer prices
Asymmetry of the
exchange rate passthrough to CPI
related to:
Threshold models (τ = threshold)
variable > τ
τ = 0.24%
Output gap (y)
0.192
∆ nominal effective
exchange rate (∆e)
Volatility of the
nominal effective
exchange rate (s)
Inflation (π)
variable ≤ τ
0.179
τ = 2.08%
0.065
0.239
τ = 4.32%
0.247
0.549
τ = level of official inflation target
0.195
0.201
Pass-through
(general)
Nonreversible linear models
t1 > t0
t1 ≤ t0
0.274
0.091
0.018
0.238
0.139
0.141
0.160
0.183
0.229
Source: Authors’ estimation.
The threshold for the output gap shows a slight deviation from zero (0.24) suggesting a small
inefficiency of the monetary policy, particularly during prosperity. In that period, a smaller
role of output gap in the explanation of inflation is combined with the bigger role of inflation
expectations. The respective coefficients are 0.14 and 0.71, compared to 0.22 and 0.49 during
the slump (table A16). It may reflect an increasing power of employees in wage negotiations
during prosperity, resulting in a higher propensity of producers to price increases. It is directly
transferred to consumers` expectations, which become more and more forward-looking.
The Wald test shows no statistically significant differences in the exchange rate pass-through
between the periods of prosperity and the periods of slump. However, there is a huge
difference in the pass-through effects between the periods of expansion (0.27) and contraction
(recession): 0.09.
Combination of these two findings would suggest that the pass-through effect may even
exceed 0.30 in the inflexion point between early and late expansion Then it drops to 0.19-0.21
in the peak and to around zero in the inflexion point between early and late recession. During
the late recession and the trough it is growing up to 0.19-0.21 and exceeds 0.30 in the early
expansion. This is coherent with the behaviour of enterprises in the business cycle,
15
conditioning their investment decisions on expected profits with a maximum in the early
expansion and a minimum in the early recession (see, e.g. Kalecki (1943) or Mankiw (2006).
The enterprises’ propensity to change prices follows profit expectations.
On the other hand, there is a strong asymmetric reaction in the periods of appreciations and
depreciations. If the zloty appreciates (or, according to the threshold model, the appreciation
rate is over 2%), the pass-through to the consumer prices declines and varies from 0.02 to
0.07. During the depreciations or slight appreciations (below 2%) and depreciations, the passthrough effect equals 0.24.
Figure 2. The cross-reaction of the asymmetric shocks of the output gap and the nominal
effective exchange rate.
3
After huge rise of the NBP's reference rate it was
consecutively cut by 9 pct. points during one year. That
rised market expectations for another changes of the
interest rates and resulted in considerable
appreciation of zloty, amplified by the drop of the
share of net exports in GDP to -2.5% in 2002. Att.:
floating exchange rate was introduced in April 2000.
2
1
Effects of over-heating. Plus
drop of the risk premium after
The EU accession.
19
97
Q
19 1
97
Q
19 3
98
Q
19 1
98
Q
19 3
99
Q
19 1
99
Q
20 3
00
Q
20 1
00
Q
20 3
01
Q
20 1
01
Q
20 3
02
Q
20 1
02
Q
20 3
03
Q
20 1
03
Q
20 3
04
Q
20 1
04
Q
20 3
05
Q
20 1
05
Q
20 3
06
Q
20 1
06
Q
20 3
07
Q
20 1
07
Q
20 3
08
Q
20 1
08
Q
3
0
-1
-2
-3
Recovery after the "Russian" crisis: low
dynamics of exports combined with high
growth of imports (share of the net export in
the GDP rises to -6,5% in 2000) resulted in
increase of the risk premium for the Polish
zloty. Plus rise of the NBP's reference rate
by 6 pct. points to combat inflation.
The EU pre-accession shock:expected rise
of prices caused high increase of demand
resulted in over-reaction of supply.
Increasing uncertainty involved a slight
depreciation of zloty.
High pass-through (over 0.25)
Average pass-through
Low pass-through (below 0.05)
Source: Authors’ estimations.
Combining the pass-through effect over the business cycle and during appreciations and
depreciations we can define periods of low and high pass-through,. There is low pass-through
if appreciation and contraction occur simultaneously. The high pass-through is characteristic
for periods of a simultaneous occurrence of expansion and depreciation. However, the
probability that such phenomenon takes place over the business cycle is relatively low. ,
Examining the data for Poland, we find two periods of high and low pass-through (Figure 2).
The first one occurred in 1998, when the Russian economy contracted due to a severe
financial crisis which burst out in this country in the last quarter of 1997. Poland, being in the
ascending phase of the business cycle, was therefore affected by a negative external shock.
During the fall of the GDP growth rate, the pass-through remained at an average level. The
coincidence of a faster growth and the depreciation, i.e. the high pass-through, occurred just
after the shock and lasted about four quarters. The rapid recovery was followed by a
16
slowdown and the appreciation of the zloty induced by the market expectations for successive
cuts in the interest rates. The period of the low pass-through effect lasted about 7 quarters and
ended up together with the slowdown of the exchange rate appreciation.
The second episode of such a coincidence occurred before Poland’s accession to the EU. The
“ordinary” recovery was accelerated by the external positive shock. An excessive rise in
demand resulted in an over-reaction of supply. Increasing uncertainty together with a
diminishing interest rates disparity involved a slight depreciation of the zloty. The period of
high pass-through effect lasted from the 2q 2003 until the 1q 2004 (Poland joined the EU in
May 2004). It was followed, similarly to the first episode, by a period of a low pass-through,
lasting till the 1q of 2005.
This can suggest that episodes of high and low pass-through occur together and immediately
after an exogenous shock. Seemingly, it does not matter whether the shock is positive or
negative.
The average volatility of the Polish zloty calculated for each quarter between 1997 and 2008
as a standard deviation of the daily data is equal to 7.1%, and varies from 2.8% in 2007 to
over 10% in 2003 and 2004 (Figure 3). The threshold value for the volatility of the nominal
effective exchange rate of the zloty was estimated at the level of 4.3%, with 29 observations
over the threshold and 17 below. The pass-through effect seems to be strongly asymmetric:
for the volatility over 4.3% it is equal to 0.25 and 0.55 otherwise. The lower pass-through in
the case of higher volatility can reflect the producers’ reluctance to frequent price changes due
to the menu costs. These results are consistent with Pollard and Coughlin (2003) or Correa
and Minella (2006).14 Generally, the periods of high volatility of the zloty are simultaneously
these of high pass-through; hence they do not raise the total effect.
Figure 3. History of the Polish zloty exchange rate
14
crowling peg to
the currency
basket
12
5,00
crowling peg to the currency
basket + crowling band (from +/7% in 1995 to +/-15% in 1999)
floating exchange rate
4,50
4,00
10
3,00
EU accession
2,50
2,00
PLN/USD
6
fixed exchange rate
in %
8
3,50
1,50
4
1,00
2
0,50
0,00
9
20
07
-1
231
20
08
-1
031
1
1
1
0
20
06
-1
22
20
05
-1
23
20
04
-1
23
20
03
-1
23
1
9
1
PLN/EUR volatility
20
02
-1
23
20
01
-1
23
20
00
-1
22
1
USD/EUR volatility
19
99
-1
23
1
1
19
98
-1
23
19
97
-1
23
9
PLN/USD volatility
19
96
-1
23
1
0
19
95
-1
22
19
94
-1
23
19
93
-1
23
19
9
19 09 01
19 0-1 -04
91 2-3
19 12- 1
92 31
-1
231
0
PLN/USD exchange rate (right scale)
Source: NBP data. Authors’ calculation.
Examining the role of the inflation environment in the exchange rate pass-through we do not
estimate the threshold – it was set up at the level of the NBP inflation target. For the
14
However Correa and Minella got different results for appreciations and depreciations.
17
nonreversible linear function models the sample was divided according to the ascending or
descending periods of inflation. The results are not clear – the Wald test on the lack of
asymmetry produces p-value at the level around 10%. Hence we decided to reject the
hypothesis of asymmetric reaction.
5. Conclusions for the monetary policy.
In the paper we do an ‘exchange rate pass-through exercise’ on the Polish data, so the results
can be country-specific. We find no evidence of a non-linear reaction of import prices with
respect to the exchange rate changes. Due to the market imperfections asymmetry arises in the
domestic links of the distribution chain. Our estimations of the consumer prices’ reaction
show that the exchange rate pass-through tends to vary alongside business cycle. It is
relatively small during contractions and high during expansions. It also differs for changes in
the nominal effective exchange rate below and above a threshold level, which is about 2
percent. In periods of increased exchange rate volatility, the pass-through effect becomes
lower and tends to rise in the time of lower volatility.
Defining periods of the high and low pass-through effect (as a combination of the passthrough over the business cycle and during appreciations and depreciations) we find that the
high pass-through occurs together with exogenous shocks and the low pass-through immediately after such shocks, no matter if they are positive or negative. Hence, during the
“ordinary” economic fluctuations, monetary policy faces average level of the pass-through
rather than an asymmetric one.
The Polish example shows that even if there is some evidence of asymmetries with respect to,
say, the output gap and appreciations and depreciations, to have a proper overview of its
potential impact on the monetary policy, all sources of possible asymmetries should be
considered.
18
Statistical Appendix
Table A1: Exchange rate pass-through. Estimation based on the McCarthy’s SVAR.
Pass-through effect
2 quarters
after →
for ↓
4 quarters
8 quarters
Est.02
Est.08
Est.02
Est.08
Est.02
Est.08
Import transaction prices (PM)
0.51
0.50
0.69
0.65
0.79
0.73
Price index of the sold
production of industry (PPI)
0.27
0.24
0.50
0.44
0.59
0.51
Consumption price index (CPI)
0.17
0.11
0.36
0.19
0.42
0.21
Source: Authors’ calculations
Table A2: Time decomposition of the pass-through effect.
Time decomposition of the passthrough effect for ↓ (total P-T=100)
Import transaction prices (PM)
Q0
Quarter after shock
Q1
Q2
Q3
17
49
25
4
5
Price index of the sold production of
industry (PPI)
12
35
29
10
14
Consumption price index (CPI)
10
42
31
7
10
Q4 to Q8
Source: Authors’ calculations
Table A3. Exchange rate pass-through for Poland.
Period of
estimation
Method applied
1998-2008
VAR/VECM
Przystupa,Wróbel
1991-2004
Ca’Zorzi et al.
1999-2004
Korhonen, and Wachtel
1998-2008
Model based on the
Phillips curve
Przystupa and Wróbel
Model NSA used at NBP
1990-1999
One-equation models
Campa and Goldberg
1998-2008
Exchange rate pass-through to:
Import prices
Consumer prices
Short-term Long- term Short- term Long- term
0.81
~ 0.70
1.30
~ 0.25
0.56
0.09
0.09
0.22
0.27
0.50
0.99
19
Table A4: Exchange rate pass-through.
Period of
estimation Method used
Euro zone
1990-2002
VAR/VECM
Faruqee
1983-2004
Ca’Zorzi et al.
Pass-through effect for:
Import prices
Consumer prices
After 2 q
After 8 q
After 2 q
After 8 q
0.42
1.17
0.01
0.02
0.55
0.72
0.06
0.13
0.66
0.81
Single eq. model
1990-2004
Campa et al.
Structural model
DIT countries
1971-1983
1984-2004
USA
1990-2002
Gagnon and
Ihrig
VAR/VECM
Faruqee
1983-2004
Ca’Zorzi et al.
0.18
0.03
0.15
0.30
0.21
0.38
0.26
0.41
0.01
0.02
Single eq. model
1990-2004
Campa et al.
Table A5. Unit root test (ADF)
Variable, in logs, nsa
Trend and intercept,
statistic, p-values in ( )
-2.450
(0.350)
ptIMP
-2.941
(0.160)
ptH
t- Intercept, t-statistic, p-values
in ( )
-2.36
(0.158)
-1.386
(0.580)
ptF
-2.237
(0.458)
0.754
(0.992)
et
-1.525
(0.806)
-1.750
(0.400)
ptIMP
-0.983
(0.936)
Trend and intercept,
statistic, p-values in ( )
-2.640
(0.266)
ptH
-3.085
(0.122)
-1.607
(0.720)
ptF
-2.756
(0.221)
0.275
(0.974)
e
NEER
t
Variable, in logs, sa
20
-0.524
(0.877)
t- Intercept, t-statistic, p-values
in ( )
-2.699
(0.082)
Table A6. Johansen cointegration test - trace statistics (NEER)
Hypothesized no
of ce(s)
None
At most 1
At most 2
At most 3
Eigenvalue
Trace statistics
0.493755
0.317904
0.207372
0.103637
60.42062
31.14904
14.69788
4.704618
0.05
critical
value
54.07904
35.19275
20.26184
9.164546
Probability
0.0219
0.1009
0.5615
0.5125
Table A7. Johansen cointegration test - trace statistics (bilateral exchange rate)
Hypothesized no Eigenvalue
Trace statistics
0.05
critical Probability
of ce(s)
value
None
0.457465
57.91189
54.07904
0.0219
At most 1
0.399639
32.22877
35.19275
0.1009
At most 2
0.161961
10.79939
20.26184
0.5615
At most 3
0.077288
3.378395
9.164546
0.5125
Table A8. Cointergating vector - NEER
VECM (cointegration Johansen), t-stat in [ ]
unrestricted
α1
α2
α3
1.44
[2.21]
0.95
[5.19]
restricted: α1 = α 2
Chi-square=0.56, p. 0.45
α1
α2
α3
-0.81
0.81
[7.50]
1.13
[17.65]
restricted: α1 = α 2 = 1
Chi-square=9.64, p. 0.008
restricted: α1 = α 2 = 1, α 3 =0
Chi-square=11.1, p. 0.011
restricted: α1 = α 2 , α 2 + α 3 =1
Chi-square=10.21, p. 0.006
-0.91
21
[5.92]
[7.50]
Table A9. Cointegrating vector, bilateral exchange rate EUR/PLN
VECM
(cointegration Fully modified least squares
Johansen), t-stat in [ ]
t-stat in [ ]
unrestricted
-0.71 [2.47]
0.68
[4.73]
α1
α2
α3
0.41
[0.35]
0.74
[1.09]
0.83
[2.27]
0.82
[4.14]
restricted: α1 = α 2
Chi-square=0.03, p. 0.86
Chi-square=0.018, p. 0.91
α1
α2
α3
-0.78
0.67
[7.15]
restricted: α1 = α 2 =1
Chi-square=7.24, p. 0.026
Chi-square=12.46, p. 0.002
restricted: α1 = α 2 = 1, α 3 =0
Chi-square=7.49, p. 0.058
Chi-square=260, p. 0.000
restricted: α1 = α 2 ,
α 2 + α 3 =1
Chi-square=6.87, p. 0.032
Chi-square=40.0, p. 0.000
[5.28]
0.78
[5.28]
0.67
[7.15]
0.71
[7.22]
0.84
[15.29]
Table A10. Test of weak exogeneity
Variable
Chi-square (1)
H
2.49
p
e
0.011
H
2.67
p
Table A11. Dynamic import price equation
Variable
Coefficient
0.0011
β0
Probability
0.11
0.92
0.102
t-stat
0.16
β1
β 2,0
-0.464
-3.02
0.51
3.94
β 4,0
0.898
1.82
Table A12. Dynamic import price equation: appreciation of the EUR/PLN, usable obs.: 24.
Variable
Coefficient
t-stat, p-value in ( )
A
-0.0023
-0.16 (0.874)
β0
β1A
-0.654
-2.96
(0.008)
A
β 2,0
0.55
1.58
(0.131)
A
β 4,0
1.29
1.8
(0.091)
22
Table A13. Dynamic import price equation: depreciation of the EUR/PLN, usable obs.: 20
Variable
Coefficient
t-stat
D
-0.0055
-0.39
β0
β1D
-0.225
-1.03
D
β 2,0
0.599
2.07
D
β 4,0
0.77
1.12
Table A14. Dynamic import price equation: positive EC , usable obs.: 21.
Variable
Coefficient
t-stat
EC +
0.00183
0.08
β0
β1EC +
-0.31
-0.63
EC +
β 2,0
0.66
4.27
EC +
β 4,0
0.003
0.0031
Table A15. Dynamic import price equation: negative EC , usable obs.: 23.
Variable
Coefficient
t-stat
EC −
-0.019
-1.62
β0
β1EC −
-0.65
-2.01
EC −
β 2,0
0.16
0.82
EC −
β 4,0
2.0
3.39
23
Table A16. Phillips curve
π tq,k = α1,qk Etπ t +1 + (1 − α1,qk − α2,q k )π t −1 + α2,q k (∆etr−1 ) + α3,q k yt −2 + ε t
i
i
i
i
i
Coefficients
α1
qi > τ
1-α1- α2
α2
α3
Threshold models (τ = threshold)
i
No. of obs.
after adj.
p-value of
J-statistic
0.708
(7.16)
0.484
-0.192
(2.50)
0.138
(1.81)
22
0.59
0.497
(4.80)
0.476
(3.98)
0.682
-0.179
(2.43)
-0.065
(5.19)
0.219
(1.79)
0.318
(2.79)
19
0.64
21
0.19
0.450
(3.26)
0.931
(2.32)
0.789
-0.239
(3.24)
-0.247
(2.63)
0.236
(2.67)
0.271
(1.75)
19
0.99
25
0.23
0.611
(3.76)
0.700
(2.88)
0.938
-0.549
(2.56)
-0.195
(3.17)
0.548
(1.81)
0.425
(2.08)
15
0.68
15
0.37
0.738
(8.01)
0.463
-0.201
(1.98)
0.064
(1.74)
26
0.46
Output gap
qi ≤ τ
qi > τ
0.589
∆NEER
qi ≤ τ
qi > τ
Volatility
NEER
qi ≤ τ
qi > τ
0.316
0.495
Inflation
qi ≤ τ
Nonreversible linear models
ti+1 > ti
0.568
(2.30)
0.706
-0.274
(1.96)
0.563
(2.11)
21
0.44
0.499
(3.78)
0.577
(3.73)
0.562
-0.091
(1.88)
-0.018
(1.63)
0.306
(1.82)
0.543
(1.68)
20
0.52
21
0.29
0.389
(1.73)
0.576
(5.21)
0.849
-0.238
(1.69)
-0.140
(2.36)
0.151
(1.61)
0.132
(1.68)
19
0.98
20
0.21
0.619
(3.34)
0.922
(2.78)
0.522
-0.141
(2.17)
-0.160
(1.72)
0.359
(1.90)
0.287
(1.98)
21
0.99
21
0.61
0.949
(5.17)
0.417
-0.183
(1.87)
0.262
(1.63)
20
0.27
Output gap
ti+1 ≤ ti
ti+1 > ti
0.441
∆NEER
ti+1 ≤ ti
ti+1 > ti
Volatility
NEER
ti+1 ≤ ti
ti+1 > ti
0.564
0.238
Inflation
ti+1 ≤ ti
24
Figure A1. Import prices (year 2000=100), overall PPI and PPI for the domestic market (year
2000=100), bilateral exchange rate, foreign prices (deflator of euro area export prices of
manufacturing, year 2000=100).
140
130
130
120
120
110
110
100
100
90
90
80
80
70
70
97
97 98 99 00 01 02 03 04 05 06 07
98
99
00
01
02
PPI
Import prices
.29
110
.28
108
.27
106
.26
104
.25
03
04
05
06
07
PPI_domestic
102
.24
100
.23
98
.22
96
.21
94
.20
97
98
99
00
01
02
03
04
05
06
97
07
98
99
00
01
Exchange rate(EUR/PLN)
02
03
04
05
06
07
Foreign prices
.03
-1.0
.02
-1.1
.01
-1.2
.00
-1.3
-.01
-1.4
-.02
-1.5
-.03
-1.6
97 98 99 00 01 02 03 04 05 06 07 08
97
Output gap
98
99
00
01
02
03
04
05
06
Nominal effective exchange rate
25
07
08
References
Alvarez R., Jaramillo P., and Selaive J. (2008). ‘Exchange rate pass-through into import
prices: The case of Chile’, Working Paper No. 465, Central Bank of Chile,
www.bcentral.cl/eng/studies/working-papers/465.htm.
Anderton B. (2003). ‘Extra-euro area manufacturing import prices and exchange rate passthrough’ ECB Working Paper No. 219, Frankfurt, European Central Bank,
www.ecb.int/pub/pdf/scpwps/ecbwp.219.pdf.
Bache I., W. (2002). ‘Empirical Modelling of Norwegian Import Prices’, Working Paper No.
1, Norges Bank, http://www.norgesbank.no/upload/import/publikasjoner/arbeidsnotater/pdf/arb-2002-01.pdf.
Bache I., W. (2007). ‘Econometrics of Exchange rate Pass-Through’, Doctoral Dissertations
in Economics No. 6, Norges Bank. www.norgesbank.no/upload/63207/bache_doctoral_dissertation.pdf.
Bailliu J., and Fujii E. (2004). ‘Exchange Rate Pass-through and the Inflation Environment in
Industrialized Countries: An Empirical Investigation’, Working Paper No. 2004-21, Bank of
Canada, www.bank-banque-canada.ca/en/res/wp/2004/wp04-21.pdf.
Barhoumi K. (2006). ‘Differences in long run exchange rate pass-through into import prices
in developing countries: An empirical investigation’, Economic Modelling, 23, pp. 926-951.
Bergin P. R., and Feenstra R. C. (2001). ‘Pricing-to-market, staggered contracts, and real
exchange rate persistence’, Journal of International Economics, 54, pp.333-359.
Bussière M. (2007). ‘Exchange rate pass-through to trade prices. The role of non-linearities
and asymmetries’, ECB Working Paper No. 822, Frankfurt, European Central Bank,
http://www.ecb.int/pub/pdf/scpwps/ecbwp.822.pdf.
Campa J. M. and Goldberg L. S. (2002). ‘Exchange Rate Pass-Through into Import Prices: A
Macro or Micro Phenomenon?’, NBER Working Paper No. 8934, Cambridge, MA: NBER,
http://www.nber.org/papers/w8934.
Campa J. M., Goldberg L. S., and Gonzalez-Minguez J.M. (2005). ‘Exchange rate PassThrough to Import Prices in the Euro Area’, NBER Working Paper No. 11632, Cambridge,
MA: NBER, http://www.nber.org/papers/w11632.
Campa J. M., Gonzalez-Minguez J.M., and Sebastiá-Barriel M. (2006). ‘Non-linear
adjustment of import prices in the European Union’ Documentos de Trabajo No. 0635,
Banco de España, www.bde.es/informes/be/docs/dt0635e.pdf.
Caner M., and Hansen B. E. (2004). ‘Instrumental variable estimation of a threshold model’,
Econometric Theory, 20, pp. 813-843.
26
Ca`Zorzi M., Hahn E., and Sanchez M. (2007). ‘Exchange rate pass-through in Emerging
Markets’, ECB Working Paper Series No. 739, Frankfurt, European Central Bank,
http://www.ecb.int/pub/pdf/scpwps/ecbwp.739.pdf.
Cholewiński R. (2008). ‘Wpływ zmian kursu walutowego na dynamikę procesów
inflacyjnych’, NBP Working Paper No. 226, National Bank of Poland,
http://www.nbp.pl/publikacje/materialy_i_studia/ms226.pdf.
Correa A. S., and Minella A. (2006). ‘Nonlinear Mechanisms of the Exchange Rate PassThrough: A Phillips curve model with threshold for Brazil’, Working Paper Series No. 122,
Banco Central do Brasil, www.bcb.gov.br/pec/wps/ingl/wps122.pdf.
Corsetti G., and Dedola L. (2002). ‘Macroeconomics of international price discrimination’,
ECB Working Paper No. 176, Frankfurt, European Central Bank,,
www.ecb.int/pub/pdf/scpwps/ecbwp.176.pdf.
Devereux M. B., Engel Ch., and Storgaard P. E. (2003). ‘Endogenous Exchange Rate PassThrough when Nominal Prices are Set in Advance’, NBER Working Paper No. 9543,
Cambridge, MA: NBER, http://www.nber.org/papers/w9543.
Devereux M. B., and Yetman J. (2002). ‘Price Setting and Exchange rate Pass-Through:
Theory and Evidence’, Working Paper No. 22/2002, Hong-Kong Institute for Monetary
Research,
www.hkimr.org/cms/upload/publication_app/pub_full_0_2_26_wp200222_text.pdf.
Dornbusch R. (1987). ‘Exchange Rates and Prices’, American Economic Review, 77, pp. 93106.
Faruqee H. (2006). ‘Exchange rate pass-through in the euro area’, IMF Staff Papers, 53,
Washington, D.C.: IMF, pp.63-88,
www.imf.org/External/Pubs/FT/staffp/2006/01/pdf/faruqee.pdf .
Filardo A. J. (1998). ‘New Evidence of the Output Cost of Fighting Inflation’, Economic
Review.83, pp.33-61, Federal Reserve Bank of Kansas City,
www.kansascityfed.org/publicat/econrev/PDF/3q98fila.pdf.
Franses P. H., and Van Dijk D. (2000). Nonlinear time series models in empirical finance,
Cambridge: Cambridge University Press.
Froot K., and Klemperer P. (1989). ‘Exchange rate pass-through when market share matters’,
American Economic Review, 79, pp. 637-654.
Fuhrer J., and Moore G. (1995). ‘Inflation Persistence’, The Quarterly Journal of Economics,
110, pp. 127-159.
Gagnon J. E., and Ihrig J. E. (2004). ‘Monetary Policy and Exchange Rate Pass-Through’,
International Journal of Finance and Economics, 9, pp.315-338.
Goldberg P. K. (1995). ‘Product differentiation and oligopoly in international markets: The
case of the US automobile industry’, Econometrica, 63, pp. 891-951.
27
Goldberg P. K., and Knetter M. M. (1997). ‘Goods price and exchange rate: what have we
learned’, Journal of Economic Literature, 35, pp. 1243-1272.
Goldfajn, I., and Werlang S. (2000). ‘The pass-through form depreciation to inflation: A panel
study’, Working Paper Series No. 5, Banco Central do Brasil
www.bcb.gov.br/pec/wps/ingl/wps05.pdf.
Hansen B. E. (2000). ‘Sample Splitting and Threshold Estimation’ Econometrica, 68(3), pp.
575-604.
Herzberg, V., Kapetanios G., and Price G. (2003).’Import prices and exchange rate passthrough: theory and evidence from the United Kingdom’, Working Paper No. 182, Bank of
England, www.bankofengland.co.uk/publications/workingpapers/wp182.pdf.
Hjalmarsson E., and Österholm P. (2007). ‘Testing for Cointegration Using the Johansen
Methodology when Variables are Near-Integrated’, IMF Working Paper No. 07/141,
Washington D.C.: IMF.
Houck J.P. (1977). ‘An Approach to Specifying and Estimating Nonreversible Functions’,
American Journal of Agricultural Economics, 59, pp. 570-572.
Ihrig J. E., Marazzi M., and Rothenberg A. D. (2006). ‘Exchange Rate Pass-Through in the
G-7 Countries’, International Finance Discussion Papers, No. 851, Board of Governors of the
Federal Reserve System.
Kalecki M. (1943). Studies in Economic Dynamics, London: George Allen and Unwin Ltd.
Kłos B., Kokoszczyński R., Łyziak T., Przystupa J., and Wróbel E. (2005). ‘Structural
Econometric Models in Forecasting Inflation at the National Bank of Poland’, NBP Working
Paper No. 31, National Bank of Poland,
www.nbp.pl/publikacje/materialy_i_studia/31_en.pdf.
Khundrakpam J. K. (2007). ‘Economic reforms and exchange rate pass-through to domestic
prices in India’, BIS Working Paper No.225, Basel, Bank for International Settlements,
http://www.bis.org/publ/work225.pdf.
Korhonen I., and Wachtel P. (2005). ‘A Note on Exchange rate Pass-Through in CIS
Countries’, BOFIT Discussion Paper No. 2/2005, Bank of Finland,
http://www.bof.fi
Łyziak T., Przystupa J., and Wróbel E. (2008). ‘Monetary Policy Transmission in Poland: A
Study of the Importance of Interest Rate and credit Channels’, SUERF Studies No.1, Vienna,
SUERF.
Mankiw N. G. (2006). Macroeconomics, New York. N.Y.: Worth Publishers, 6th edition.
Mann C.L. (1986). ‘Prices, Profit Margins, and Exchange Rates’, Federal Reserve Bulletin,
72, pp. 366-379.
28
McCarthy J. (2000). ‘Pass-through of exchange rates and import prices to domestic inflation
in some industrialized countries’, Federal Reserve Bank of New York Staff Reports No. 111,
New York, N.Y., http://www.newyorkfed.org/research/staff_reports/sr111.html.
Pollard P. S., and Coughlin C. C. (2003). ‘Size Matters: Asymmetric Exchange Rate PassThrough at the Industrial Level’, Working Paper No. 2003-029C, Federal Reserve Bank of St.
Louis, http://research.stlouisfed.org/wp/2003/2003-029.pdf.
Posedel P., and Tica J. (2007). ‘Threshold Autoregressive Model of Exchange Rate PassThrough Effect: The Case of Croatia’, Working Paper No. 07-15, University of Zagreb,
Faculty of Economics and Business, http://web.efzg.hr/repec/pdf/Clanak%2007-15.pdf.
Przystupa J. (2002). ‘The exchange rate in the monetary transmission mechanism’, NBP
Working Paper No. 25, National Bank of Poland,
www.nbp.pl/publikacje/materialy_i_studia/25_en.pdf.
Przystupa J., and Wróbel E. (2006). ‘Looking for an Optimal Monetary Policy Rule: The Case
of Poland under IT Framework’, NBP Working Paper No. 38, National Bank of Poland,
www.nbp.pl/publikacje/materialy_i_studia/38_en.pdf.
Rogoff K. (1996). ‘The Purchasing Power Parity Puzzle’, Journal of Economic Literature, 34,
pp.647-668.
Silver M. (2007). ‘Do Unit Value Export, Import and Terms of Trade Represent or
Misrepresent Price Indices?’, Working Paper, No. 07/121, Washington, D.C.: IMF,
www.imf.org/external/pubs/ft/wp/2007/wp07121.pdf.
Taylor J. B. (1993). ‘Discretion versus policy rules in practice’, Carnegie-Rochester
Conference Series on Public Policy, Elsevier, 39, pp. 195-214.
Taylor J. B. (2000). ‘Low inflation, pass-through, and the pricing power of firms’, European
Economic Review, 44, pp. 1389-1408.
Tong H. (1990). Non-linear time series: a dynamical system approach, Oxford: Oxford
University Press,.
Wickremasinghe G., and Silvapulle P. (2004). ‘Exchange Rate Pass-Through to Manufactured
Import Prices: The Case of Japan’, International Trade, No. 0406006.
Woodford M. (2003). Interest and Prices: Foundations of a Theory of Monetary Policy,
Princeton N.J.: Princeton University Press.
Wolffram R. (1971). ‘Positive Measures of Aggregate Supply Elasticities: Some New
Approach–Some Critical Notes’, American Journal of Agricultural Economics, 53, pp. 356359.
29